Optimized automated cardiac MR scar quantification with GAN‐based data augmentation

• Published in: Computer methods and programs in biomedicine 2022-11, Vol.226, p.107116-107116, Article 107116
• Main Authors: Lustermans, Didier R.P.R.M., Amirrajab, Sina, Veta, Mitko, Breeuwer, Marcel, Scannell, Cian M.
• Format: Article
• Language: English
• Subjects: Cardiac MRI; Deep learning; Generative adversarial networks; Myocardial scar quantification; Synthetic data
• ISSN: 0169-2607; 1872-7565
• DOI: 10.1016/j.cmpb.2022.107116

•Accurate automated scar quantification from late gadolinium-enhancement cardiac MRI using deep learning.•Cascaded pipeline comprised of sequential bounding box detection, myocardium segmentation, and scar segmentation outperforms the direct segmentation of scar.•The inclusion of synthetic data, generated with a GAN, during training improves performance. The clinical utility of late gadolinium enhancement (LGE) cardiac MRI is limited by the lack of standardization, and time-consuming postprocessing. In this work, we tested the hypothesis that a cascaded deep learning pipeline trained with augmentation by synthetically generated data would improve model accuracy and robustness for automated scar quantification. A cascaded pipeline consisting of three consecutive neural networks is proposed, starting with a bounding box regression network to identify a region of interest around the left ventricular (LV) myocardium. Two further nnU-Net models are then used to segment the myocardium and, if present, scar. The models were trained on the data from the EMIDEC challenge, supplemented with an extensive synthetic dataset generated with a conditional GAN. The cascaded pipeline significantly outperformed a single nnU-Net directly segmenting both the myocardium (mean Dice similarity coefficient (DSC) (standard deviation (SD)): 0.84 (0.09) vs 0.63 (0.20), p